Secure data transfer over a network

RELATED APPLICATIONS

This application is related to and shares a common disclosure with commonly-assigned copending applications U.S. application Ser. No. 10/791,415, titled “System and Method for Performing Security Operations on Network Data” and U.S. application Ser. No. 10/791,239, titled “System and Method for Performing Cryptographic Operations on Network Data”, both filed on Mar. 2, 2004, the entire contents of which are incorporated herein by reference.

BACKGROUND

Today, most data transfers sent over the public networks, such as the Internet, are left unprotected against attacks. Even users of private networks that rely on public network communication facilities to connect end-user terminals and workstations in the private network to servers and other terminals in the public network are vulnerable to attacks. Moreover, recent industry studies have found that over half of all private network security breaches originated from within the private network. The situation results mainly because popular packet communication protocols, such as TCP/IP, do not have protection mechanisms designed into their protocol stacks. Consequently, any terminal connected to a TCP/IP network can intercept, replay, or produce IP packets sent over the network.

In response to the situation, the Internet Engineering Task Force (or IETF) defined Internet Protocol Security (or IPSec) to provide encryption-based security in TCP/IP networks. IPSec is a network-layer (e.g., the IP layer of TCP/IP) security framework that provides end-to-end network security services such as authentication, data integrity, confidentiality (or encryption), and anti-replay protection for IPv4 and IPv6 data sent over public and private networks. IPSec is defined in several IETF publications, including RFC 2401, titled “Security Architecture for the Internet Protocol”, and RFC 2411, titled “IP Security Document Roadmap”.

The first IPSec solutions were implemented primarily using software. While these solutions provided acceptable performance at the time of their introduction, software implementations of IPSec can no longer keep pace with the rapidly improving bandwidth of today's workstations, servers, and routers. Some hardware-assisted IPSec solutions have been introduced that provide improved performance over prior software implementations, but these systems still do not deliver the high-bandwidth performance needed to support today's communication systems. Moreover, these early hardware-assisted IPSec solutions generally employ pipelined, flow-through processor architectures, the performance of which can be impacted by “bursty” traffic flows.

In addition, many IPSec solutions focus on performing security operations, such as encryption and authentication, and either ignore or place a lesser emphasis on higher workload functions such as compression and quality-of-service concerns, including flow policy, congestion control, and traffic shaping.

SUMMARY

Accordingly, a system and method are disclosed for secure data transfer over a network. According to an exemplary embodiment a system for secure data transfer over a network includes memory and a memory controller configured to transfer data received from the network to the memory. The system includes a processor, having logic configured to retrieve a portion of the data from the memory using the memory controller. The processor also includes logic configured to perform security operations on the retrieved portion of the data, and logic configured to store the operated-on portion of the data in the memory using the memory controller. The memory controller is further configured to transfer the operated-on portion of the data from the memory to the network.

According to another exemplary embodiment, a method for secure data transfer over a network is disclosed that includes transferring data from the network to memory using a memory controller. A portion of the data is retrieved from the memory into a processor using the memory controller. Security operations are performed on the retrieved portion of the data using the processor. The operated-on portion of the data is stored in the memory and transferred m the memory to the network using the memory controller using the memory controller.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings provide visual representations which will be used to more fully describe the representative embodiments disclosed here and can be used by those skilled in the art to better understand them and their inherent advantages. In these drawings, like reference numerals identify corresponding elements, and:

<figref idrefs="DRAWINGS">FIG. 1</figref> illustrates a system for secure data transfer over a network, according to an exemplary embodiment;

<figref idrefs="DRAWINGS">FIG. 2</figref> is a flowchart illustrating a method for secure data transfer over a network using the system shown in <figref idrefs="DRAWINGS">FIG. 1</figref>, according to an exemplary embodiment;

<figref idrefs="DRAWINGS">FIG. 3</figref> illustrates a system for performing security operations on network data, according to an exemplary embodiment;

<figref idrefs="DRAWINGS">FIG. 4</figref> illustrates an IPSec process for performing security operations using the system of <figref idrefs="DRAWINGS">FIG. 3</figref>, according to an exemplary embodiment;

<figref idrefs="DRAWINGS">FIG. 5</figref> is a flowchart illustrating a method for performing security operations on network data using the system of <figref idrefs="DRAWINGS">FIG. 3</figref>, according to an exemplary embodiment;

<figref idrefs="DRAWINGS">FIG. 6</figref> illustrates a system for performing cryptographic operations on network data, according to an exemplary embodiment; and

<figref idrefs="DRAWINGS">FIG. 7</figref> is a flowchart illustrating a method for performing cryptographic operations on network data using the system of <figref idrefs="DRAWINGS">FIG. 6</figref>, according to an exemplary embodiment.

DETAILED DESCRIPTION

Various aspects will now be described in connection with exemplary embodiments, including certain aspects described in terms of sequences of actions that can be performed by elements of a computer system. For example, it will be recognized that in each of the embodiments, the various actions can be performed by specialized circuits or circuitry (e.g., discrete and/or integrated logic gates interconnected to perform a specialized function), by program instructions being executed by one or more processors, or by a combination of both.

Thus, the various aspects can be embodied in many different forms, and all such forms are contemplated to be within the scope of what is described. For each of the various aspects, any such form of embodiment can be referred to here as “logic configured to” perform, or “logic that” performs a described action.

A system for secure data transfer over a network according to an exemplary embodiment is shown in <figref idrefs="DRAWINGS">FIG. 1</figref>. The system shown can be described as a “store-and-forward” system in contrast to a “flow-through” system. In a store-and-forward system, received data, such as a data packet, is first stored in memory, after which the data is operated on by task-specific processors. When processing is complete, the data is forwarded from the memory to its destination. In a flow-through system, the received data is passed (or flows) through a series of task-specific processors that operate on the data as it flows through the system. In a flow-through system, portions of the received data, e.g., the packet header, can be operated-on and transmitted to its destination before remaining portions of the data, e.g., the packet payload, has been operated-on or even received. The processors in flow-through systems typically operate in a pipelined manner.

The system of <figref idrefs="DRAWINGS">FIG. 1</figref> includes memory <b>102</b>, such as the Double Data Rate (DDR) Dynamic Random Access Memory (DRAM), or DDR-DRAM, shown in the figure. Any type of large, high-density, high-latency storage can be used as the memory <b>102</b>. The memory <b>102</b> is preferably homogenous, e.g., of a same type, to simplify memory allocation during the various tasks performed by the system. The memory <b>102</b> can be allocated in blocks, e.g., of 512 K buffers, each of which blocks can be partitioned into a number of memory banks (not shown).

The system of <figref idrefs="DRAWINGS">FIG. 1</figref> includes a memory controller <b>104</b> configured to transfer data received from the network (not shown) to the memory <b>102</b>. The memory controller <b>104</b> can include logic configured to dynamically allocate additional memory blocks <b>102</b> depending on the system load. For example, additional memory blocks <b>102</b> can be dynamically added when the data (or traffic) being processed by the system peaks, or becomes “bursty”. An advantage of store-and-forward systems, such as that shown in <figref idrefs="DRAWINGS">FIG. 1</figref>, is that the overall performance (or throughput) of the system can be maintained under “bursty” conditions by dynamically adding memory to store the additional traffic data. Although the latency through the system can be impacted, the system's performance under “bursty” conditions can outperform comparable flow-through systems.

According to an exemplary embodiment, the memory controller <b>104</b> can include logic configured to reference the number of memory banks in a sequence that minimizes a memory access time. The memory controller <b>104</b> can also include logic configured to include a request to reference the memory into one of a group of read requests and a group of write requests. Additional logic in the controller <b>104</b> can be configured to execute all requests included in one of the groups of read requests and write requests before executing a request included in the other group. These functions of the memory controller <b>104</b> are enhanced through the use of a common, homogeneous memory, such as the DDR-DRAM <b>102</b> shown in <figref idrefs="DRAWINGS">FIG. 1</figref>. This can lead to reduced memory access times and more efficient memory allocation than can be possible in systems that use various task-specific memories, such as Contents Addressable Memories (CAMs) and internal Static RAMs (or SRAMs). Additional logic can be configured in the system, e.g. in the controller <b>104</b>, to include error correction code with the data transferred to or stored in the memory <b>102</b>. Logic can also be included to detect and correct errors in the data retrieved or transferred from the memory based on the error correction code included with the data.

The system of <figref idrefs="DRAWINGS">FIG. 1</figref> also includes a processor, such as the Embedded Processor Complex (EPC) <b>106</b> shown in the figure. EPC, as used here, denotes a processing structure having multiple processors, e.g., protocol processors, together with dedicated hardware-assist coprocessors, for performing various tasks. The protocol processors can be “pico-processors” capable of executing a limited set of instructions. As will be understood by those skilled in the art, the term “pico-processor” is used here to denote a processor having an instruction set that is somewhat more limited than that of a typical Reduced Instruction Set Computer (or RISC processor). Although, an EPC is illustrated and described as the processor <b>106</b> in <figref idrefs="DRAWINGS">FIG. 1</figref>, other types of processing structures can be used without deviating greatly from what is described below.

The processor <b>106</b> includes logic configured to retrieve a portion of the data from the memory <b>102</b> using the memory controller <b>104</b>. As discussed above, the processing system shown in <figref idrefs="DRAWINGS">FIG. 1</figref> is a store-and-forward system. Accordingly, after the data is transferred from the network (not shown) to the memory (e.g., the “store” step), the data is moved to the EPC <b>106</b> via the memory controller <b>104</b> for processing. The processor <b>106</b> also includes logic configured to perform security operations on the retrieved portion of the data. For example, the processor <b>106</b> can include logic configured to execute programmable instructions (e.g., pico-code) for implementing IPSec. As discussed above, IPSec can provide end-to-end security services such as authentication, data integrity, confidentiality (or encryption), and anti-replay protection for IPv4 and IPv6 data over public and private networks.

To provide encryption services, the logic configured to perform security operations can include both logic configured to obscure the portion of the data when the retrieved portion is non-secure data, and logic configured to decipher the portion of the data when the retrieved portion is secure data. As used here, non-secure data can include unencrypted (or clear text) data exchanged within a private or trusted network, and secure data can include encrypted data exchanged within a public or untrusted network.

The logic configured to obscure and decipher the portion of data can include pico-code to implement IP Encapsulating Security Payload (ESP). ESP is described in detail in RFC 2406 published by the IETF. In addition, the logic can be configured to perform any of several industry-standard cryptographic algorithms for bulk encryption supported by ESP, including the Data Encryption Standard (DES), the Triple-DES (3DES), and the Advanced Encryption Standard (AES) algorithms. These algorithms are discussed in greater detail below in conjunction with the exemplary cryptographic coprocessor shown in <figref idrefs="DRAWINGS">FIG. 6</figref>.

To provide authentication services, the logic configured to perform security operations can include logic configured to determine an integrity of the portion of data. The logic configured to determine an integrity of the portion of data can include pico-code to implement IP Authentication Header (or AH), as described in detail in IETF publication RFC 2402, and pico-code to implement ESP authentication.

The logic can also be configured to perform any of the several industry-standard hashing algorithms, such as Secure Hash Algorithm (SHA-1) and Message Data 5 (MD5), each compatible with AH and ESP. The hashing algorithms can be used to determine a message authentication code (MAC) associated with the portion of the data. The MAC can then be compared with an expected value to determine the integrity of the portion of the data. The hashing algorithms described above are discussed in greater detail below in conjunction with the exemplary cryptographic coprocessor shown in <figref idrefs="DRAWINGS">FIG. 6</figref>.

The processor <b>106</b> also includes logic configured to store the operated-on portion of the data in the memory using the memory controller <b>104</b>. As used here, the “operated-on” portion refers to the portion of data after the security operations have been performed. The memory controller <b>104</b> is further configured to transfer the operated-on portion of the data from the memory <b>102</b> to the network (not shown). Accordingly, with the store-and-forward arrangement shown in <figref idrefs="DRAWINGS">FIG. 1</figref>, data is stored into the memory <b>102</b>, operated on by the EPC <b>106</b>, returned to the memory <b>102</b> after being operated-on by the EPC <b>106</b>, and then the operated-on portion is forwarded from the memory <b>102</b> to the network (not shown) where it is routed to its destination node.

To aid in the transfer of data to and from the network, the system of <figref idrefs="DRAWINGS">FIG. 1</figref> can include a network interface coupled to the memory controller <b>104</b>. The network interface can include a first data moving unit (DMU) <b>108</b> configured to exchange e.g., via the data buses <b>112</b>. secure data with a first portion of the network (not shown). The first network portion can be a public or untrusted portion of the network, such as the Internet. The network interface can also include a second DMU <b>110</b> configured to exchange, e.g., via the data buses <b>114</b>. non-secure data with a second portion of the network (not shown). The second network portion can be a private or trusted portion of the network, such as a corporate Intranet.

The arrangement described above can be used to perform security operations on data received by the first DMU <b>108</b> from the first portion of the network, and to forward the operated-on data to the second portion of the network via the second DMU <b>110</b>. The dotted arrows <b>122</b> shown in <figref idrefs="DRAWINGS">FIG. 1</figref> depict a logical path of the data through the system under this mode of operation. The arrangement described above can also be used to perform security operations on data received by the second DMU <b>110</b> from the second portion of the network, and to forward the operated-on data to the first portion of the network via the first DMU <b>108</b>. The dotted arrows <b>124</b> shown in <figref idrefs="DRAWINGS">FIG. 1</figref> depict a logical path of the data through the system in this mode of operation. As will be understood by those skilled in the art, the system of <figref idrefs="DRAWINGS">FIG. 1</figref> can be further configured, such that data received from the network portion coupled to one of the DMUs <b>108</b>, <b>110</b>, can be operated on, and then forwarded to the same network portion using the one DMU <b>108</b>, <b>110</b>. A combination of the dotted arrows <b>122</b>, <b>124</b> depicts the possible logical paths of the data through the system under these modes of operation.

Each DMU <b>108</b>, <b>110</b>, can be configured for both Ethernet and POS transmission, and can be capable of moving data at 1 Gb/s when receiving and transmitting data in either modes. When configured for Ethernet, each DMU <b>108</b>, <b>110</b> can support either one port of 1 Gigabit Ethernet or ten ports of Fast Ethernet (e.g., at 10/100 Mb/s). To support 1 Gigabit Ethernet, each DMU <b>108</b>, <b>110</b> can be configured as either a gigabit media-independent interface (GMII) or a ten-bit interface (TBI). To support Fast Ethernet, each DMU <b>108</b><b>110</b> can be configured as a serial media-independent interface (SMII) supporting ten Ethernet ports. When configured for POS mode, each DMU <b>108</b>, <b>110</b> can support both clear-channel and channelized optical carrier (OC) interfaces.

The network interface can also include a first serializer/deserializer (SERDES) circuit <b>116</b> coupled between the first DMU <b>108</b> and the first network portion (not shown), and a second SERDES <b>118</b> coupled between the second DMU <b>110</b> and the second network portion (not shown). Each SERDES can be configured to convert serial data received from the respective network portions to a parallel format, and to convert parallel data received from the respective DMUs to a serial format. The design of such SERDES circuitry, which is known to those skilled in the art, is beyond the scope of this document and will not be discussed in detail.

According to an exemplary embodiment, the processor <b>106</b> can include logic configured to perform quality-of-service (QoS) operations on the data in coordination with performing the security operations. QoS operations can provide improved service to selected network traffic over various technologies, such as over IP-routed Ethernet networks. A goal of QoS is to provide priority including dedicated bandwidth, controlled jitter and latency (required by some real-time and interactive traffic), and improved loss characteristics the selected network traffic.

The logic configured to perform QoS operations can include logic configured to identify an information flow associated with the portion of the data. An information flow can be defined by combination of source and destination network addresses, a combination of source and destination network socket numbers, by a session identifier, or as any portion of data (e.g., or packet) from a certain application or from an incoming interface in the network. The logic configured to identify an information flow can include pico-code to implement Differentiated Services (Diffserv), as described in detail in IETF publications RFC 2474 and 2475. For example, the portion of the data can be marked by setting bits in an IP header field (e.g., the DS Field) of an IPv4 or IPv6 packet to identify the information flow associated with the portion. Alternatively, an information flow associated with portion of the data can be identified on a “per-hop basis”, e.g., based on a device in the network that the portion of data is being carried on.

Once a flow is identified, logic can be configured in the processor <b>106</b> to determine a priority of the information flow. Additional logic in the processor <b>106</b> can be configured to schedule at least one of the retrieving the portion of the data and the transferring the operated-on portion of the data from memory <b>102</b> based on the priority of the information flow associated with the portion of the data. For example, portions of the data associated with a higher priority information flow can be retrieved from the memory <b>102</b> and operated on by the EPC <b>106</b> before portions having a lower priority are operated on.

In addition, the operated-on portions of the data associated with a higher priority information flow can be transferred from the memory <b>102</b> to the network before transferring out of the memory <b>102</b> portions of the data associated with a lower priority flow. In this way, QoS policies associated with the data operated on by the EPC <b>106</b> can be enforced. For example, the scheduler <b>120</b> shown in <figref idrefs="DRAWINGS">FIG. 1</figref> can be used to enforce QoS policies by scheduling the transfer of data from the memory <b>102</b> to the network via the memory controller <b>104</b>. The scheduler <b>120</b> can also be used for traffic shaping (e.g., queuing)) and policing (e.g., discarding) portions of data associated with a particular information flow.

Logic in the processor <b>106</b> can be configured to decipher the portion of the data prior to the identifying of the information flow when the retrieved portion is secure data (e.g., encrypted). Likewise, the logic can be configured to obscure the portion of the data after the identifying of the information flow when the retrieved portion is non-secure data (e.g., unencrypted). Coordinating the performing of the QoS and security operations in this manner can vastly improve the QoS of traffic operated on by the security processor <b>106</b> by enforcing QoS policies that can not be enforced on encrypted data.

According to an exemplary embodiment, the processor <b>106</b> can also include logic configured to compress the portion of the data using the processor prior to performing the security operations when the retrieved portion is non-secure data, and logic configured to decompress the portion of the data in the processor after performing the security operations when the retrieved portion is secure data. For example, the processor <b>106</b> can include pico-code to implement IP Payload Compression (IPComp) as described in detail in IETF publication RFC 2393.

IPComp is a protocol that can reduce the size of IP datagrams, and can be particularly useful when encryption is applied to the IP datagrams. Encrypting the IP datagram can causes the data to be random in nature, rendering compression at lower protocol layers (e.g., PPP Compression Control Protocol [RFC-1962]) ineffective. If both compression and encryption are required, compression must be applied before encryption, and decompression applied after decryption. The processor <b>106</b> can also include additional logic configured to implement at least one of a Lempel-Ziv-Stac (LZS) and an Adaptive Lossless Data Compression (ALDC) compression algorithm. The LZS compression algorithm is described in detail in IETF publication RFC 2395. ALDC is described in Standard QIC-154, Revision A, Mar. 10, 1994, and in ECMA-222, ISO/IEC 15200, and ANSI x3.280-1996. Both the LZS and ALDC compression algorithms can be applied to provide lossless compression to the payloads of IP datagrams including, but not limited to, those that adhere to the IPComp protocol.

<figref idrefs="DRAWINGS">FIG. 2</figref> is a flowchart illustrating a method for secure data transfer over a network. The method can be carried out using a store-and-forward arrangement such as that shown in <figref idrefs="DRAWINGS">FIG. 1</figref>. In block <b>202</b>, data from the network is transferred to memory, e.g., DDR-DRAM <b>102</b>, using a memory controller, such as memory controller <b>104</b>. In block <b>204</b>, a portion of the data is retrieved from the memory <b>102</b> into a processor, such as the EPC <b>106</b>, using the memory controller <b>104</b>. In block <b>206</b>, security operations are performed on the retrieved portion of the data using the processor <b>106</b>. The security operations can include IPSec authentication, data integrity, confidentiality (or encryption), and anti-replay protection, as described above in conjunction with the arrangement of <figref idrefs="DRAWINGS">FIG. 1</figref>. Next, in block <b>208</b>, the operated-on portion of the data is stored in the memory <b>102</b> using the memory controller <b>104</b>. Finally, in block <b>210</b>, the operated-on portion of the data is transferred from the memory <b>102</b> to the network using the memory controller.

According to an exemplary embodiment, the security operations can include at least one of obscuring the portion of the data when the retrieved portion is non-secure data; deciphering the portion of the data when the retrieved portion is secure data; and determining an integrity of the portion of data. Algorithms for encryption and authentication for use with IPSec are discussed above, and are described below in detail in conjunction with the cryptographic processor shown in <figref idrefs="DRAWINGS">FIG. 6</figref>.

In addition, QoS operations can be performed on the data in coordination with performing the security operations using the processor <b>106</b>. As discussed above, the QoS operations can include identifying an information flow associated with the portion of the data; determining a priority of the information flow; and scheduling at least one of the retrieving the portion of the data and the transferring the operated-on portion of the data from memory based on the priority of the information flow associated with the portion of the data. The QoS and security operations can be coordinated, such that the portion of the data deciphered prior to the identifying of the information flow when the retrieved portion is secure data, and the portion obscured after the identifying of the information flow when the retrieved portion is non-secure data. Additional shaping (queuing) and policing (discarding) of portions of the data associated with a particular information can performed.

According to an exemplary embodiment, the portion of the data can be compressed using the processor <b>106</b> prior to performing the security operations when the retrieved portion is non-secure data, and the portion can be decompressed in the processor <b>106</b> after performing the security operations when the retrieved portion is secure data. For example, the IPComp protocol and/or at least one of the LZS and ALDC compression algorithms can be applied to provide lossless compression of IP datagrams as discussed above.

According to yet another exemplary embodiment, a request to reference the memory can be included into one of a group of read requests and a group of write requests. All pending requests included in one of the groups of read requests and write requests can be executed before executing the requests included in the other group. Persons skilled in the art will understand that grouping read/write requests can provide for efficient memory access by reducing bus set-up and reconfiguration times.

In a related embodiment, the executing of all requests included in one of the groups of read requests and write requests can occur when a sum of the requests included in one of the groups corresponds to a predetermined amount of the memory. The predetermined amount can correspond to an amount of the memory <b>102</b> (e.g., an optimal number of memory banks) that maximizes memory throughput. In addition, portions of the memory <b>102</b> (e.g., a number of memory banks) can be referencing in a sequence that minimizes a memory access time. Finally, error correction code can be included with the data transferred to or stored in the memory <b>102</b> to detect and possibly correct errors in the data retrieved or transferred from the memory <b>102</b>.

The instructions of a computer program as illustrated in <figref idrefs="DRAWINGS">FIG. 2</figref> for secure data transfer over a network can be embodied in any computer readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer based system, processor containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions.

As used here, a “computer readable medium” can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer readable medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium. More specific examples (a non exhaustive list) of the computer readable medium can include the following: an electrical connection having one or more wires, a portable computer diskette, a random access memory (RAM), a read only memory (ROM), an erasable programmable read only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read only memory (CDROM).

A system for performing security operations on network data according to an exemplary embodiment is shown in <figref idrefs="DRAWINGS">FIG. 3</figref>. The system shown in <figref idrefs="DRAWINGS">FIG. 3</figref> corresponds to the EPC <b>106</b> shown in <figref idrefs="DRAWINGS">FIG. 1</figref>. The system includes multiple processors, e.g., protocol processors <b>304</b>, together with dedicated hardware-assist coprocessors, e.g., crypto coprocessors <b>306</b>, for performing various tasks. The term coprocessor is used here to describe a processor configured to assist another processor in a particular task. A coprocessor need not be of lesser complexity or different in any other way to its corresponding processor, although often coprocessors can include hardware specially designed to optimally perform a given task. In contrast, a processor can be more general-purpose in design than its coprocessor, and can be configured to perform tasks defined by executable instructions, such as pico-code.

The system shown in <figref idrefs="DRAWINGS">FIG. 3</figref> includes memory, such as the shared memory pool <b>302</b> shown in the figure. The shared memory pool can comprise SRAM embedded with the logic of EPC <b>106</b>, although other types of memory can be used. The system also includes a data coprocessor (e.g., the data store interface, or DS IF) configured to transfer data into and out of the memory <b>302</b>, e.g., to the DDR-DRAM <b>102</b> of <figref idrefs="DRAWINGS">FIG. 1</figref> via bus <b>310</b>.

As briefly discussed above, the system includes a plurality of processors <b>304</b> coupled to the memory <b>302</b> and to the data coprocessor DS IF. Each processor <b>304</b> can be configured to perform, in parallel to one another, security operations on a portion of the data. The processors <b>304</b> can be pico-processors capable of executing a limited set of instructions for implementing a various number of security-related protocols, such as IPSec, IPComp, DiffServ, and the like. The system also includes a plurality of security coprocessors (e.g., cryptographic, or crypto coprocessors) <b>306</b> coupled to the memory <b>302</b>. Each security coprocessor <b>306</b> is coupled to a respective one of the processors <b>304</b>, and is configured to assist the respective processor <b>304</b> in performing security operations on the portion of the data.

<figref idrefs="DRAWINGS">FIG. 4</figref> illustrates an IPSec process for performing security operations using the system of <figref idrefs="DRAWINGS">FIG. 3</figref>, according to an exemplary embodiment. The IPSec process is referred to as a “bump-in-the-wire” (BITW) implementation. The BITW process can be implemented in a security processor gateway node arranged between two networks to provide security operations on data passed between the networks. The figure also depicts an exemplary apportionment of tasks between the processors <b>304</b> and the security coprocessors <b>306</b> in performing the security operations.

For example, <figref idrefs="DRAWINGS">FIG. 4</figref> shows that each of the plurality of processors <b>304</b> can include logic configured to identify a security association related to the portion of the data, such as “SA” Lookup Engine pico-code <b>402</b> shown in the figure. A security association (or SA) can be an instance of security policy and keying material applied to the data included in an information flow. IPSec SAs are unidirectional and they are unique in each security protocol, e.g., ESP and AH. A set of SAs are needed for a protected data pipe, one per direction per protocol. For example, if you have a pipe that supports ESP between peers, one ESP SA is required for each direction. SAs can be uniquely identified by destination (IPSec endpoint) address, security protocol (AH or ESP), and a security parameter index (SPI), used to distinguish between SAs having a same endpoint address and protocol. Information regarding SAs can be found in IETF publication RFC 2408, titled “Internet Security Association and Key Management Protocol (ISAKMP)”.

Each of the plurality of processors <b>304</b> can also include logic configured to filter the portion of the data based on the identified SA. For example, the BITW process of <figref idrefs="DRAWINGS">FIG. 4</figref> includes Packet Filter Engine pico-code <b>404</b>. The Packet Filter Engine can be used to discard packets passing between a public/untrusted network and a private/trusted network that lack the necessary security credentials. The processors <b>304</b> can also include logic configured to divide the portion of the data into fragments and to reassemble the fragments into the portion. For example, pico-code blocks <b>406</b> for performing fragmentation and reassembly of fragments are included in the BITW process as shown in <figref idrefs="DRAWINGS">FIG. 4</figref>. Fragmentation can be required when the portion of the data exceeds a specified data (or packet) length. As shown in the figure, fragmentation/reassembly of the portion of data should occur after/before the performing of the security operations on the data. Each processor <b>304</b> can also include logic configured to identify a sequence associated with the portion of the data. For example, the sequence ID can be used by the Anti-replay Control pico-code <b>408</b> to reject or discard replayed packets to provide a form of partial sequence data integrity.

Each security coprocessor <b>306</b> can include logic configured to obscure the portion of the data when the portion is non-secure data and logic configured to decipher the portion of the data when the portion is secure data. For example, the security coprocessors <b>306</b> can include logic configured to perform encryption/decryption of the portion of the data using one of the DES, 3DES, or AES standard encryption algorithms discussed above. The security coprocessors <b>306</b> can also include logic configured to determine an integrity of the portion of the data. For example, the security coprocessors <b>306</b> can include logic configured to perform authentication of the portion of the data using one of the SHA-1 or MD5 standard hashing algorithms discussed above.

In addition to the above, each security coprocessor <b>306</b> can include logic configured to establish an SA related to the portion of the data. The SA can include information used to obscure and decipher the portion and to determine the integrity of the portion. For example, each security coprocessor <b>306</b> can include a Public Key Engine (PKE) <b>414</b> that can include logic used to assist in the generation of keying material, e.g., cryptographic keys generated according to the Internet Key Exchange (IKE) hybrid protocol, as defined in IETF publication RFC 2409. The keying material can be used to establish an SA related to the portion of data and to authenticate and encrypt the data, e.g., using AH and ESP. The IKE hybrid protocol uses the Diffie-Hellman method of establishing shared keys over an insecure medium.

The system of <figref idrefs="DRAWINGS">FIG. 3</figref> can also include a search engine coprocessor <b>308</b> coupled to the memory <b>302</b> and to the plurality of processors <b>304</b>. The search engine coprocessor <b>308</b> can be configured to exchange control information between at least one of the memory and external system memory, such as the DDR-DRAM <b>102</b> shown in <figref idrefs="DRAWINGS">FIG. 1</figref> via bus <b>312</b>, and each of the plurality of processors <b>304</b> for use in performing security operations on the data. The control information can include tables, counters, and other data used by the processors <b>304</b> in implementing the various security protocols.

The search engine coprocessor <b>308</b> can support various tree search algorithms, including searching Fixed Match Trees (e.g., fixed size patterns requiring an exact match, such as Layer 2 Ethernet media access control tables), Longest Prefix Match Trees (e.g., variable length patterns requiring variable length matches, such as subnet IP forwarding), and Software Managed Trees (e.g., two patterns defining either a range or a bit mask set, such as the masks used for filter rules). The search engine coprocessor <b>308</b> can provide for complex searching of the memory <b>302</b> and the external memory <b>102</b> for control information without having to interrupt execution of the protocol processors <b>304</b>.

According to an exemplary embodiment, the system of <figref idrefs="DRAWINGS">FIG. 3</figref> also includes a memory coprocessor (e.g., a control access bus interface or CAB IF) coupled to the plurality of processors <b>304</b>, the memory <b>302</b>, and the external system memory <b>102</b>. The memory coprocessor CAB IF can be configured to determine a status of the memory <b>302</b> and the external system memory <b>102</b>.

Each of the plurality of processors <b>304</b> can be further configured to perform, in parallel to one another, QoS operations on the portion of the data in coordination with performing the security operations. As discussed above, the logic configured to perform QoS operations can include logic configured to identify an information flow associated with the portion of the data using DiffServ. Once a flow is identified, logic can be configured in the processors <b>304</b> to determine a priority of the information flow. Additional logic in the processors <b>304</b> can be configured to manage the transfer of data into and out of the memory <b>302</b> based on the priority of the information flow associated with the data. For example, portions of the data associated with a higher priority information flow can be retrieved from the memory <b>302</b> and operated on by the processors <b>304</b> before portions having a lower priority are operated on.

In addition, the operated-on portions of the data associated with a higher priority information flow can be transferred from the memory <b>302</b>, e.g., to the external system memory <b>102</b> via the data coprocessor DS IF and bus <b>310</b>, before transferring out of the memory <b>302</b> portions of the data associated with a lower priority flow. The system can include an enqueue coprocessor coupled to the plurality of processors <b>304</b> and to the data coprocessor DS IF. The enqueue coprocessor EnQ can be configured to manage the information flow associated with the data external to the system. For example, the enqueue coprocessor EnQ can interface with the scheduler <b>120</b> of <figref idrefs="DRAWINGS">FIG. 1</figref> to schedule the transfer of data from the external system memory <b>102</b> to the network to maintain the information flow associated with the data.

A policy coprocessor Policy can be configured to assist the plurality of processors <b>304</b> in managing the transfer of the data into and out of the memory <b>302</b> by enforcing policies of the information flow associated with the data. Also, a counter coprocessor Counter can be configured to provide statistics related to the transfer of the data into and out of the memory and the enforcing of policies of the information flow.

According to an exemplary embodiment, each of the plurality of processors is configured to execute programmable instructions for performing the security operations on the portion of the data from a plurality of independent instruction streams (or threads). Each of the processors can switch between instruction steams in a single clock cycle. Switching can occur in one of the processors <b>304</b> when the processor <b>304</b> is waiting for a shared resource, e.g., the respective security processor <b>306</b> or the shared memory pool <b>302</b>. Enabling the processors <b>304</b> to switch between independent instruction streams can allow the shared resources to remain optimally utilized. Each of the plurality of security processors <b>306</b> can include separate queues corresponding to each of the independent instruction streams of the processors <b>304</b>.

The processors <b>304</b> can also include logic configured to compress the portion of the data using the processor prior to performing the security operations when the retrieved portion is non-secure data, and logic configured to decompress the portion of the data in the processor after performing the security operations when the retrieved portion is secure data. For example, the processor <b>106</b> can include pico-code to implement IPComp as discussed above. Each security processor can be configured to assist the respective processor <b>304</b> in compressing and decompressing the portion of the data. For example, the security processors <b>306</b> can include logic configured to implement at least one of the LZS and ALDC compression algorithms as discussed above.

<figref idrefs="DRAWINGS">FIG. 5</figref> is a flowchart illustrating a method for performing security operations on network data. The method can be carried out using an embedded processor complex such as that shown in <figref idrefs="DRAWINGS">FIG. 3</figref>. In block <b>502</b>, data is transferred into memory, e.g., the shared memory pool <b>302</b> of <figref idrefs="DRAWINGS">FIG. 3</figref>. In block <b>504</b>, security operations are performed on respective portions of the data in parallel using a plurality of processors, e.g., the protocol processors <b>304</b>. The security operations can include IPSec authentication, data integrity, confidentiality (or encryption), and anti-replay protection, as described above in conjunction with the arrangements of <figref idrefs="DRAWINGS">FIGS. 1 and 3</figref>. In block <b>506</b>, a plurality of security coprocessors, e.g., the crypto coprocessors <b>306</b>, are used to assist in performing the security operations on the respective portions of the data. Each security coprocessor <b>306</b> is coupled to a respective one of the processors <b>304</b>. Finally, in block <b>508</b>, the operated-on portions of the data are transferred out of the memory <b>302</b>.

According to an exemplary embodiment, the security operations performed by each of the processors <b>304</b> can include identifying an SA related to a portion of the data; filtering the portion of the data based on the identified SA; dividing the portion of the data into fragments; reassembling the fragments into the portion of data; and identifying a sequence associated with the portion of the data. For example, an IPSec process for performing security operations using the system of <figref idrefs="DRAWINGS">FIG. 3</figref> is illustrated in <figref idrefs="DRAWINGS">FIG. 4</figref>.

According to another exemplary embodiment, the security operations assisted by each of the security coprocessors <b>306</b> can include obscuring a portion of the data when the portion is non-secure data; deciphering the portion of the data when the portion is secure data; determining an integrity of the portion of the data; and establishing an SA related to the portion of the data, wherein the security association includes information used in obscuring and deciphering the portion and in determining the integrity of the portion. Algorithms for encryption and authentication for use with IPSec are discussed above, and are described below in detail in conjunction with the cryptographic processor shown in <figref idrefs="DRAWINGS">FIG. 6</figref>. Control information can be exchanged between at least one of the memory <b>302</b> and external system memory <b>102</b> and each of the plurality of processors <b>304</b> for use in performing security operations on the data.

In another exemplary embodiment, QoS operations are performed on the respective portions of the data in parallel using the plurality of processors <b>304</b> in coordination with performing the security operations. The QoS operations performed by each of the processors <b>304</b> can include identifying an information flow associated with the data; determining a priority of the information flow; and managing the transfer of data into and out of the memory <b>302</b> based on the priority of the information flow associated with the data.

In related embodiments, the information flow can be managed after transferring the operated-on portions of the data associated with the information flow out of the memory <b>302</b>, e.g., by using the enqueue coprocessor EnQ of <figref idrefs="DRAWINGS">FIG. 3</figref> to interface with the scheduler <b>120</b> of <figref idrefs="DRAWINGS">FIG. 1</figref>. Policies of the information flow associated with the data can be enforced, e.g., using the policy coprocessor Policy of <figref idrefs="DRAWINGS">FIG. 3</figref>. In addition, statistics related to the transfer of the data into and out of the memory <b>302</b> and the enforcing of policies of the information flow can be provided, e.g., via the counter coprocessor Counter of <figref idrefs="DRAWINGS">FIG. 3</figref>.

Respective portions of the data can be compressed prior to performing the security operations when the portions are non-secure data, and decompressing after performing the security operations when the portions are secure data. For example, pico-code for implementing the IPComp protocol can be included in each of the processors <b>304</b> to perform the compression/decompression of the data. In addition, each security processor can be used to assist the respective processor in compressing and decompressing the portions of the data, e.g., via logic configured to implement at least one of the LZS and ALDC compression algorithms on IPComp-compliant datagrams.

The instructions of a computer program as illustrated in <figref idrefs="DRAWINGS">FIG. 5</figref> for performing security operations on network data can be embodied in any computer readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer based system, processor containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions.

A system for performing cryptographic operations on network data according to an exemplary embodiment is shown in <figref idrefs="DRAWINGS">FIG. 6</figref>. The system shown in <figref idrefs="DRAWINGS">FIG. 6</figref> corresponds to one of the cryptographic processors <b>306</b> shown in <figref idrefs="DRAWINGS">FIG. 3</figref>. The system includes an input interface <b>602</b> configured to receive data into the system, e.g., from the shared memory pool <b>302</b> shown in <figref idrefs="DRAWINGS">FIG. 3</figref>. A plurality of processors <b>604</b>, <b>606</b>, <b>608</b> are included in a cascaded arrangement. The cascaded arrangement can be a sequential arrangement in which the output of one processor is coupled to the input of at least one other processor in the arrangement.

As shown in <figref idrefs="DRAWINGS">FIG. 6</figref>, each processor <b>604</b>, <b>606</b>, <b>608</b> in the arrangement has an input <b>1</b> coupled to the input interface <b>602</b>, and an output O coupled to respective inputs of each of the other processors downstream in the arrangement. The processors <b>604</b>, <b>606</b>, <b>608</b> are each configured to perform respective cryptographic operations, e.g., compression, encryption, authentication, and the like, on the data. An output interface <b>610</b> is coupled to the input interface <b>602</b>, and to the output O of each of the processors <b>604</b>, <b>606</b>, <b>608</b>. The output interface <b>610</b> is configured to transmit data out of the system, e.g., to the shared memory pool <b>302</b>, and to direct the data through the system in coordination with the input interface <b>602</b> according to a predetermined algorithm. As used here, “downstream” refers to a progression of the processors <b>604</b>, <b>606</b>, <b>608</b> from the input interface <b>602</b> to the output interface <b>610</b>. The system can include logic configured to determine a checksum associated with the data transmitted out of the system.

According to an exemplary embodiment, a first processor <b>604</b> has its data inputs <b>1</b> coupled only to the input interface <b>602</b>. Consequently, the processor <b>604</b> is the first processor in the cascaded arrangement. The first processor <b>604</b> can be configured to compress uncompressed data and to decompress compressed data. For example, in <figref idrefs="DRAWINGS">FIG. 6</figref>, the first processor <b>604</b> is configured to perform at least one of the LZS and ALDC compression algorithms on the data. The data to be operated on can be, but need not be, compliant with the IPComp compression protocol.

A second processor <b>606</b> in the arrangement has a first input <b>1</b> coupled to the input interface <b>602</b> and a second input <b>2</b> coupled to an output <b>0</b> of the first processor <b>604</b>. The second processor <b>606</b> can be configured to obscure non-secure data and to decipher secure data. For example, the second processor can be configured to obscure and decipher the data using at least one of the DES, 3DES, and AES algorithms. DES is defined by the National Institute of Standards and Technology (NIST) in the Federal Information Processing Standards Publications (FIPS PUBS) 46-2. 3DES is defined in the American National Standard (ANS) working draft X9.52-1998. AES is defined in NIST FIPS PUBS 197. The second processor <b>606</b> can be configured to perform other encryption algorithms if necessary.

A third processor <b>608</b> in the arrangement has a first input <b>1</b>\coupled to the input interface, a second input <b>2</b> coupled to an output O of the first processor <b>602</b>, and a third input <b>3</b> coupled to an output O of the second processor <b>606</b>. The third processor <b>608</b> is configured to determine an integrity of the data. For example, the third processor can be configured to determine the integrity by hashing the data according to at least one of a SHA-1 and an MD5 algorithm. Hashing takes an input message (or data) of arbitrary length and produces a fixed length digest (or MAC) that is unique for the given message. SHA-1 (or HMAC-SHA-1-96) is defined in NIST FIPS PUBS 180-1. MD5 (or HMAC-MD5-96) is defined in IETF publication RFC 1321.

Consider an example in which the predetermined algorithm requires that both compression and encryption and/or authentication be performed on a data flow. Recall that IPComp/LZS/ALDC compression must be performed prior to performing security operations on the data when the data is non-secure and that decompression must occur after performing the security operations when the data is secure. If the data in the flow is non-secure data, the data can be directed from the input interface <b>602</b> to the first processor <b>604</b> where the data is compressed. Next, the compressed data can be directed to the second processor <b>606</b> for encryption and/or to the third processor <b>608</b> for computation of a MAC. The operated-on data and computed MAC can be directed to the output interface <b>610</b>, which can then be transferred to the shared memory pool <b>302</b> for further processing or storage.

If, instead, the data in the flow is secure data, the data can be directed from the input interface <b>602</b> to the second processor <b>606</b> for decryption and/or to the third processor <b>608</b> for computation of a MAC. The operated-on data and computed MAC can then be directed to the output interface <b>610</b>, which can then be transferred to the shared memory pool <b>302</b> for further processing or storage. The input interface <b>602</b> can then retrieve the operated-on data from the shared memory pool <b>302</b> and direct the data to the first processor <b>604</b> where it is decompressed. The decompressed data can be directed to the output interface <b>610</b> for transfer back to the shared memory pool <b>302</b>.

According to an exemplary embodiment, the predetermined algorithm can be based on control information included in an SA related to the data. The input interface <b>602</b> can be configured to receive the control information and to forward the control information to each of the processors <b>604</b>, <b>606</b>, <b>608</b> for use in performing the respective cryptographic operations on the data. The control information can be gathered from an SA database (or SAD) by one of the processors <b>304</b> and stored in the shared memory pool <b>302</b> for use by the corresponding cryptographic processor <b>306</b>.

The control information can include at least one of an identity of an authentication algorithm used to hash the data, e.g., SHA-1 or MD5; an identity of an encryption algorithm used to obscure and decipher the data, e.g., DES or 3DES; keying material used by at least one of the authentication and encryption algorithms, e.g., cryptographic keys generated by the Public Key Engine <b>414</b>; and a lifetime of the SA related to the data. A IPSec mode, e.g., tunnel or transport, can also be included in the control information.

<figref idrefs="DRAWINGS">FIG. 7</figref> is a flowchart illustrating a method for performing cryptographic operations on network data. The method can be carried out using cryptographic processor such as that shown in <figref idrefs="DRAWINGS">FIG. 6</figref>. In block <b>702</b>, data is received, e.g., by the input interface <b>602</b> from the shared memory pool <b>302</b>. In block <b>704</b>, the received data is directed through a cascaded arrangement of processors, e.g., processors <b>604</b>, <b>606</b>, <b>608</b>, according to a predetermined algorithm. Each processor <b>604</b>, <b>606</b>, <b>608</b> has an input <b>1</b> coupled to the received data and an output coupled to respective inputs of each of the other processors downstream in the arrangement.

In block <b>706</b> of the method, respective cryptographic operations defined by the predetermined algorithm are performed on the received data using the plurality of processors <b>604</b>, <b>606</b>, <b>608</b>. The operated-on data is then transmitted in block <b>708</b> after performing the cryptographic operations defined by the predetermined algorithm. A checksum associated with the transmitted data can be determined. The checksum can be a 16-bit one's compliment checksum, such as a TCP/IP checksum.

According to an exemplary embodiment, uncompressed received data can be compressed and compressed received data can be decompressed using a first processor <b>604</b> in the arrangement having its data inputs <b>1</b> coupled only to the received data. The received data can be compressed and decompressed according to at least one of the LZS and ALDC compression algorithms as described above. The method can also include obscuring non-secure data and deciphering secure data using a second processor <b>606</b> in the arrangement having a first input <b>1</b> coupled to the received data and a second input <b>2</b> coupled to an output of the first processor <b>604</b>. The data can be obscured and deciphered according to at least one of the DES, 3DES and AES algorithms.

An integrity of the data can be determined using a third processor <b>608</b> in the arrangement having a first input <b>1</b> coupled to the received data, a second input <b>2</b> coupled to an output of the first processor <b>604</b>, and a third input <b>3</b> coupled to an output of the second processor <b>606</b>. The data can be hashed to determine the integrity according to at least one of a SHA-1 and an MD5 algorithm.

The predetermined algorithm can based on control information included in an SA related to the received data. The control information can be forwarded to each of the processors <b>604</b>, <b>606</b>, <b>608</b> for use in performing the respective cryptographic operations on the data. The control information can include at least one of an identity of an authentication algorithm used to hash the data; an identity of an encryption algorithm used to obscure and decipher the data; keying material used by at least one of the authentication and encryption algorithms; and a lifetime of the SA related to the data.

The instructions of a computer program as illustrated in <figref idrefs="DRAWINGS">FIG. 7</figref> for performing cryptographic operations on network data can be embodied in any computer readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer based system, processor containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions.

It will be appreciated by those of ordinary skill in the art that the concepts and techniques described here can be embodied in various specific forms without departing from the essential characteristics thereof. The presently disclosed embodiments are considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the appended claims, rather than the foregoing description, and all changes that come within the meaning and range of equivalence thereof are intended to be embraced.